proc-macros: rename parameter names

[deflator-hacker]

Decoupling rust parameter names from (de)serialized json is essential for me and useful in general.

Some usecases include:

• supporting parameter names that are reserved keywords in rust, such as type (currently impossible)
• talking to case sensitive RPC servers only accepting non-snake-case parameters (currently impossible without clippy complaining)

Those are both usecases I personally have.

The PR implements this by introducing a new attribute for the rpc macro called argument.

• documentation has been updated
• tests have been written

I previously responded to an outdated similar PR here.

Minimal working example:

#[rpc(client, server)]
pub trait MyRpc {
	#[method(name = "foo")]
	async fn foo(
		&self,
		param_a: u8,
		#[argument(rename = "type")]
		r#type: String
	) -> RpcResult<u16>;
}

I divided the changes into multiple commits because they are a bit lengthy.
This is my first open source contribution, I am eager to learn, let me know what can be improved.

[niklasad1]

Looks overall good, nice job

I'm not following why you need to rename parameters when those passed as Array (these are omitted by the client)? In such scenarios then variable doesn't correspond to the actual param name, so why is it needed?

The contrary applies to JSON Objects where the key is the same name passed in the call with some glue to convert snake_case to camelCase. Then I follow why these need to be renamed to support reserved rust keywords etc....

I'll take a closer look tomorrow.

[deflator-hacker]

I'm not following why you need to rename parameters when those passed as Array (these are omitted by the client)? In such scenarios then variable doesn't correspond to the actual param name, so why is it needed?

You are absolutely right, there is no point (actually no way) to rename parameters in arrays.
I included a check for array params in proc_macros/tests/ui/correct/rename_fn_params.rs just because I wanted to assert that I did not break anything deserialization-related by accident.

[niklasad1]

Looks great, I'll can push a fix to unblock the UI tests tomorrow it's probably the latest rustc release that broke them.

[niklasad1]

This is needed because arg_pat below can't be reserved rust names such as type

[niklasad1]

Nice work looks good

[jsdw]

LGTM, just a couple of nits!